A Kalman Filter Based Approach for Outlier Detection in Sensor Networks

Outliers are common in data collection applications with wireless sensor networks, which consist of a large number of sensor nodes, embedded in physical space. The limited power supplies and noisy sensor data put challenges for outlier detection and cleaning in sensor networks. In this paper, we propose utilizing spatial and temporal dependencies that exist sensory readings. Our approach is based on Kalman filter and we design the state transition module and measuring module of the Kalman filter to exploit the temporal and spatial dependencies of sensor data respectively. The experimental results illustrate the effectiveness of our approach.